Databricks turns Genie into an executive workhorse as rivals press for revenue clarity

Updated November 12, 2026 3:54 pm

Databricks updated its Genie AI on Wednesday so enterprise customers can push through longer, multi-step tasks using company data without constant supervision. The company is positioning the tool as an always-on decision aid for business leaders at the same time rivals are questioning how much of Databricks’ rapid growth reflects core software demand versus AI pass-through revenue.

What changed in Genie?

The headline update, reported by TechCrunch, is that Genie can now handle extended, multi-step work unsupervised. Databricks says the system uses a built-in ontology to pull context from customer data already stored on its platform, which is designed to make the agent more useful inside enterprise workflows than a generic chatbot.

In practical terms, Databricks is not just marketing Genie as a text generator. It is trying to make the product useful for tasks that require memory, context and follow-through across several steps — the sort of work that sits closer to analysis and operations than to one-off prompts.

How Databricks demonstrated it

Rather than stage a polished product demo, CEO Ali Ghodsi showed Genie using his own internal workflows, according to TechCrunch. That included having the system analyze weak points in an employee satisfaction survey, then helping him evaluate and reject proposed fixes for low scores.

He also used Genie to summarize a failed sales pitch and surface the buyer requirements that the team had missed. In another example, Genie prepared him for a podcast interview by reviewing past appearances and predicting questions.

Ghodsi said he relies on the system to surface negative information he may not hear directly as CEO, a reminder that the product is being aimed at leaders who may want an outside view of their own organizations. TechCrunch also reported that he said he uses Genie in his personal life.

“Anyone who has more than 150 employees, which is a lot of companies, a lot of people can get value out of this”

He framed that threshold with a reference to Dunbar’s number, adding, “Have you heard of Dunbar’s number? The magic number is 150”.

Why it matters

The update matters because Databricks is trying to move AI agents from novelty to utility inside the enterprise. If Genie can reliably manage multi-step tasks with company context, it could become more than a search layer or assistant and instead function as a persistent business aide for managers, sales teams and executives.

That is a meaningful shift for customers already storing data on the Databricks platform. It also speaks to a broader race in enterprise AI: vendors are trying to justify higher spending by tying models and agents to work that is measurable, repeatable and embedded in existing workflows.

For readers and buyers, the question is less whether AI can answer questions and more whether it can save time, catch blind spots or reduce the cost of routine analysis. Databricks is betting that the answer becomes clearer when the agent has access to the customer’s own data and can keep working without close supervision.

Who is Genie for?

Genie is aimed at companies large enough to have layers of information that no single executive sees in full. Ghodsi said companies with more than about 150 employees can get value from AI agents, suggesting that the tool becomes more useful once internal communication is too complex for any leader to track personally.

That helps explain why Databricks is marketing Genie as an always-on agent for business leaders. The pitch is not only about automating a task, but about creating a continuous layer of observation across data, meetings, customer conversations and performance signals.

Background: Databricks’ larger AI push

Databricks is known for selling data and AI infrastructure that enterprises use to store, process and analyze information. AI agents fit naturally into that stack because they can do more than generate text; they can execute multi-step work, use tools and draw on data access.

That context matters because Genie is being built on top of customer data already inside Databricks’ environment. According to the company, the built-in ontology provides the context needed to make the agent more relevant to each enterprise’s internal operations.

Databricks is also adding controls around how AI gets used. TechCrunch reported that Unity Gateway can cap agent spend and switch to cheaper open-weight models, a sign that the company knows cost discipline is central to enterprise adoption.

The message is straightforward: if companies are going to trust agents with longer workflows, they will want visibility into costs and the option to route requests in more economical ways.

What rivals are pressing Databricks on

While Databricks was showing off Genie, BW Businessworld reported that Snowflake CEO Sridhar Ramaswamy was pressing the company on a different issue: whether its reported growth is being inflated by AI inference revenue that passes through the business rather than reflecting pure software demand.

Ramaswamy challenged Databricks on revenue mix, cash flow, margins and dilution, and asked whether growth would slow if pass-through inference revenue were removed. That line of questioning goes to the heart of how investors should interpret fast-growing AI infrastructure companies right now.

Ghodsi responded by arguing that Databricks and Snowflake can be complementary rather than direct enemies. According to BW Businessworld, he urged Snowflake to source AI model tokens through Databricks, a sign that Databricks wants to be seen not merely as a rival platform but as a supplier inside a wider ecosystem.

How Databricks is defending its economics

BW Businessworld reported that Ghodsi said Databricks customers cut token spending by about 30% while usage rose. If accurate, that would suggest the company is using efficiency gains to drive more adoption rather than simply extracting more spend from the same activity.

The same report said Databricks disclosed an annualized revenue run-rate above $7 billion in August, said its second-quarter year-on-year growth was more than 80%, and reported that the Lakehouse business crossed a $1.5 billion revenue run-rate while Lakebase exceeded a $100 million revenue run-rate.

Those figures are important, but they need to be read carefully. A revenue run-rate is an annualized snapshot, not necessarily a forecast of full-year revenue. In the same way, year-on-year growth can be driven by several lines of business at once, including any AI-related usage that is billed through the platform.

Why it matters for the market

The debate around Databricks is bigger than one product launch. Investors and customers are trying to figure out whether the current AI boom is producing durable software economics or merely routing spend through infrastructure vendors that sit between model providers and enterprise users.

That is why Ramaswamy’s questions about revenue mix, margins and dilution matter. They are not just competitive sparring. They speak to whether a company’s growth is driven by repeatable demand for its platform or by a temporary surge in AI-related usage that may be harder to sustain.

Databricks’ August fundraising only sharpened the spotlight. BW Businessworld said the company raised $5 billion at a $190 billion valuation, placing it among the most closely watched private AI infrastructure companies in the market.

If the company can show that Genie and other AI products deepen customer reliance on the platform without crushing margins, it strengthens the case for the valuation. If not, critics will keep pressing for more detail on how much revenue is core software and how much is effectively AI plumbing.

How the numbers stack up

The figures attached to Databricks this year show both scale and scrutiny. The company is big enough to make bold claims about enterprise AI, but it is also large enough that competitors and investors now want granular proof.

• 12,000 employees — the workforce size Ghodsi cited when discussing how leaders lose direct visibility inside larger companies.
• 150 employees — the rough threshold he said marks when CEOs can no longer know everyone directly.
• 30% reduction — the token-spend cut Ghodsi said customers have achieved while usage increased.
• $7 billion — Databricks’ annualized revenue run-rate disclosed in August.
• $190 billion — the valuation BW Businessworld said Databricks reached after its August funding round.

What happens next

The next phase for Databricks is likely to be less about the reveal and more about proof. The company will need to show whether Genie is actually being used in production by enough enterprise customers to justify the “always-on” framing.

Several questions remain open. How much of Databricks’ revenue growth comes from AI inference versus core data products? Will the company publish the detailed metrics Ramaswamy asked for? Can customers independently verify the claimed token savings? And how broadly will enterprise teams trust an agent with unsupervised multi-step work?

Those questions matter because the market is no longer satisfied with demos. Buyers want evidence that AI agents reduce work without creating hidden costs, and investors want a clearer read on whether rapid expansion is coming from sustainable software demand.

For now, Databricks is making a simple bet: if Genie can help executives find what they are missing inside their own organizations, then the product becomes far more than a chatbot. It becomes a layer of decision support tied directly to the company’s data, spending controls and daily operations.

Timeline

The conversation around Databricks has moved quickly over the past few months, with product updates, growth disclosures and public competition all feeding into the same debate.

That sequence shows the company moving on two tracks at once: building a more capable enterprise agent while defending the economics behind its growth story. The outcome will depend on whether customers see Genie as a real productivity layer — and whether the market accepts Databricks’ version of what is driving the numbers.